Heterogeneous Computing

The recent proliferation of computing devices and the contexts in which they are used demand diversity in collaborative applications as well. The objective of our research is to enable conferees to share applications that are adapted to their (heterogeneous) computing environments. This is achieved by using eXtensible Markup Language (XML) for the communication medium. The conferees share the same data or a subset of that data, represented in XML, but they may see it displayed in different ways as needed or desired. The framework adapts to the computing and network environment and transforms information so that it matches the client's local capabilities and resources.

With the proliferation of mobile computing devices there is an increasing demand for applications supporting collaboration among users working in the field and in the office. A key component for collaboration in this domain is sharing and manipulation of information using very different devices and communications. We propose a novel, data-centric collaboration paradigm, where each user can obtain a subset of the shared data and the data may be visualized differently for different users. The data amount and the visualization technique reflect the user's interests and/or computing and communication capabilities. The users collaborate on and exchange data, and the data is dynamically transformed to adapt to the particular computing/network platform. The resulting design is simple yet very powerful and scalable. It is implemented and tested by developing several complex groupware applications.

The recent proliferation of computing devices and the contexts in which they are used demand diversity in collaborative applications as well. The objective of our research is to enable conferees to share applications that are adapted to their (heterogeneous) computing environments. This is achieved by using eXtensible Markup Language (XML) for the communication medium. The conferees share the same data or a subset of that data, represented in XML, but they may see it displayed in different ways as needed or desired. The framework adapts to the computing and network environment and transforms information so that it matches the client's local capabilities and resources.

The recent proliferation of computing devices and the contexts in which they are used demand diversity in collaborative applications as well. The objective of our research is to enable conferees to share applications that are adapted to their (heterogeneous) computing environments. This is achieved by using eXtensible Markup Language (XML) for the communication medium. The conferees share the same data or a subset of that data, represented in XML, but they may see it displayed in different ways as needed or desired. The framework adapts to the computing and network environment and transforms information so that it matches the client's local capabilities and resources.

With the proliferation of mobile computing devices there is an increasing demand for applications supporting collaboration among users working in the field and in the office. A key component for collaboration in this domain is sharing and manipulation of information using very different devices and communications. We propose a novel, data-centric collaboration paradigm, where each user can obtain a subset of the shared data and the data may be visualized differently for different users. The data amount and the visualization technique reflect the user's interests and/or computing and communication capabilities. The users collaborate on and exchange data, and the data is dynamically transformed to adapt to the particular computing/network platform. The resulting design is simple yet very powerful and scalable. It is implemented and tested by developing several complex groupware applications.

The recent proliferation of computing devices and the contexts in which they are used demand diversity in collaborative applications as well. The objective of our research is to enable conferees to share applications that are adapted to their (heterogeneous) computing environments. This is achieved by using eXtensible Markup Language (XML) for the communication medium. The conferees share the same data or a subset of that data, represented in XML, but they may see it displayed in different ways as needed or desired. The framework adapts to the computing and network environment and transforms information so that it matches the client's local capabilities and resources.

The recent proliferation of computing devices and use contexts demand equivalent diversity in collaborative applications. Our work on the DISCIPLE and Manifold frameworks supports the development of collaborative applications for these heterogeneous environments. Using eXtensible Markup Language (XML) for the communication medium provides for the heterogeneity. Collaborators share the same data or a subset of the data, represented in XML, but view very different displays depending on their computing capabilities. Mobile users, for example, may view a more limited budget presentation than their collaborating colleague in the office. The Manifold framework transforms information to match the client's local capabilities and resources, yet maintains important content information. We used DISCIPLE and Manifold to implement a 3D to 2D collaborating environment and measured team performance characteristics in a simple object placement task.

This final deliverable describes the main objectives, solutions, results, and lessons learned by the FP6 project -Wireless Accessible Sensor Populations‖ (WASP) that ran from September 2006 to August 2010. The WASP project explored and developed solutions to ease deployment of Wireless Sensor Networks (WSN) and their integration in enterprise systems. The deliverable presents a summary of the WASP integrated system and the application and system requirements architecture. The key hardware & software components, tools to ease deployment, (where possible) advance with respect to state-of-the-art and major lessons learned and useful for future improvements. Besides all the hard integration work, also quite some exploratory work has been conducted that will be illustrated as well. A brief overview of the dissemination and exploitation activities as well as future opportunities is described as well.

The well-known Smith-Waterman (SW) algorithm is a high-sensitivity method for local alignments. However, SW is expensive in terms of both execution time and memory usage, which makes it impractical in many applications. Some heuristics are possible but at the expense of loosing sensitivity. Fortunately, previous research have shown that new computing platforms such as GPUs and FPGAs are able to accelerate SW and achieve impressive speedups. In this paper we have explored SW acceleration on a heterogeneous platform equipped with an Intel Xeon Phi co-processor. Our evaluation, using the well-known Swiss-Prot database as a benchmark, has shown that a hybrid CPU-Phi heterogeneous system is able to achieve competitive performance (62.6 GCUPS), even with moderate lowlevel optimisations.

Several evolutionary algorithms (EAs) applied to a wide class of communication network design problems modelled under the generalized Steiner problem (GSP) are evaluated. In order to provide a fault-tolerant design, a solution to this problem consists of a preset number of independent paths linking each pair of potentially communicating terminal nodes. This usually requires considering intermediate non-terminal nodes (Steiner nodes), which are used to ensure path redundancy, while trying to minimize the overall cost. The GSP is an NP-hard problem for which few algorithms have been proposed. This article presents a comparative study of pure and hybrid EAs applied to the GSP, codified over MALLBA, a general purpose library for combinatorial optimization. The algorithms were tested on several GSPs, and asset efficient numerical results are reported for both serial and distributed models of the evaluated algorithms.

We present in this work a wide spectrum of results on analyzing the behavior of parallel heuristics (both pure and hybrid) for solving optimization problems. We focus on several evolutionary algorithms as well as on simulated annealing. Our goal is to offer a first study on the possible changes in the search mechanics that the algorithms suffer when shifting from a LAN network to a WAN environment. We will address six optimization tasks of considerable complexity. The results show that, despite their expected slower execution time, the WAN versions of our algorithms consistently solve the problems. We report also some interesting results in which WAN algorithms outperform LAN ones. Those results are further extended to analyze the behavior of the heuristics in WAN with a larger number of processors and different connectivities.

We present in this work a wide spectrum of results on analyzing the behavior of parallel heuristics (both pure and hybrid) for solving optimization problems. We focus on several evolutionary algorithms as well as on simulated annealing. Our goal is to offer a first study on the possible changes in the search mechanics that the algorithms suffer when shifting from a LAN network to a WAN environment. We will address six optimization tasks of considerable complexity. The results show that, despite their expected slower execution time, the WAN versions of our algorithms consistently solve the problems. We report also some interesting results in which WAN algorithms outperform LAN ones. Those results are further extended to analyze the behavior of the heuristics in WAN with a larger number of processors and different connectivities.

This article presents an empirical study devoted to characterize the computational efficiency behavior of an evolutionary algorithm (usually called canonical) as a C program. The study analyzes the effects of several implementation decisions on the execution time of the resulting evolutionary algorithm. The implementation decisions studied include: memory utilization (using dynamic vs. static variables and local vs. global variables), methods for ordering the population, code substitution mechanisms, and the routines for generating pseudorandom numbers within the evolutionary algorithm. The results obtained in the experimental analysis allow us to conclude that significant improvements in efficiency can be gained by applying simple guidelines to best program an evolutionary algorithm in C.

This work presents a novel parallel micro evolutionary algorithm for scheduling tasks in distributed heterogeneous computing and grid environments. The scheduling problem in heterogeneous environments is NP-hard, so a significant effort has been made in order to develop an efficient method to provide good schedules in reduced execution times. The parallel micro evolutionary algorithm is implemented using MALLBA, a general-purpose library for combinatorial optimization. Efficient numerical results are reported in the experimental analysis performed on both well-known problem instances and large instances that model medium-sized grid environments. The comparative study of traditional methods and evolutionary algorithms shows that the parallel micro evolutionary algorithm achieves a high problem solving efficacy, outperforming previous results already reported in the related literature, and also showing a good scalability behavior when facing high dimension problem instances.

Scheduling is a capital problem when using distributed heterogeneous computing (HC) and grid environments to solve complex problems. The scheduling problem in heterogeneous environments is NP-hard, so a significant effort has been made to develop efficient methods for solving the problem. However, few works have faced realistic grid-sized problem instances. This work presents a parallel CHC (pCHC) evolutionary algorithm codified over MALLBA, a general-purpose library for combinatorial optimization, for solving the scheduling problem in HC and grid environments. Efficient numerical results are reported in the experimental analysis performed on both a standard benchmark and a set of large-sized problem instances specially designed in this work. The comparative study shows that pCHC is able to achieve high problem solving efficacy, significantly improving over traditional deterministic scheduling methods, while also showing a good scalability behavior when solving large problem instances.

This paper analyzes some technical and practical issues concerning the heterogeneous execution of parallel genetic algorithms (PGAs). In order to cope with a plethora of different operating systems, security restrictions, and other problems associated to multi-platform execution, we use Java to implement a distributed PGA model. The distributed PGA runs at the same time on different machines linked by different kinds of communication networks. This algorithm benefits from the computational resources offered by modern LANs and by Internet, therefore allowing researchers to solve more difficult problems by using a large set of available machines. We analyze the way in which such heterogeneous systems affect the genetic search for two problems. Our conclusion is that super-linear performance can be achieved not only in homogeneous but also in heterogeneous clusters of machines. In addition, we study some special features of the running platforms for PGAs, and basically find out that heterogeneous computing can be as efficient or even more efficient than homogeneous computing for parallel heuristics. # 2002 Elsevier Science (USA)

An ad hoc grid is a heterogeneous computing and communication system without a fixed infrastructure; all of its components are mobile. Energy management is a major concern in an ad hoc grid. One important aspect of energy management is to minimize the energy consumption during a mission. In an ad hoc grid, communication and computations are deeply intertwined, and any energy optimization must consider both types of activities together rather than separately. The mapping (defined as matching and scheduling) of tasks onto machines with varied computational capabilities has been shown, in general, to be an NP-complete problem. Therefore, heuristic techniques are required to efficiently map tasks to machines in an ad hoc grid so as to minimize the energy consumed due to communication and computation. This research evaluates and compares energy management issues for resource allocation in ad hoc grids using six static heuristics.

Abstract. MyGrid is a complete grid solution for running Bag-of-Tasks applications (ie parallel applications whose tasks are independent) over whatever resources are available to the user. MyGrid middleware empowers users to interoperate with heterogeneous computational resources across geographic and administrative boundaries. Due to MyGrid's flexible architecture, it is easy to add a new kind of machine in the grid through an abstraction called grid machine. The grid machines implementations currently available in ...

An important problem in heterogeneous computing (HC) is predicting task execution time. A methodology is introduced for determining the execution time distribution for a given data parallel program that is to be executed in an SIMD, MIMD (SPMD), and/or mixed-mode SIMD/MIMD (SPMD) HC environment. The program is assumed to contain operations and constructs whose execution times depend on input-data values. The methodology uses a block-based approach to transform the program into a flow analysis tree and computes the execution time distribution for the program, given the execution modes for each node in the flow analysis tree, an estimated execution time distribution for each operation in both modes, and appropriate probabilistic models for control and data conditional constructs. The results are directly applicable to both mixed-machine and mixed-mode HC systems.

In a heterogeneous computing (HC) environment consisting of different types of machines, an application program is decomposed into subtasks, each of which is computationally homogeneous. The goal is to execute subtasks on the machines in such a way that the total program execution time is minimized. A mathematical framework is presented that models the matching of subtasks to machines, scheduling of subtasks' computation, scheduling of intermachine communication steps, and selection of sources of shared data items for intermachine communication (data relocation). The goal of this work is to generate a provably optimal scheme for communicating shared data among subtasks as an enhancement to any given matching and scheduling. Initially, it is assumed that at any instant in time, only one machine is being used for program execution and only one subtask is being executed. Based on this assumption, a polynomial algorithm is introduced to optimize scheduling and data relocation with respect to any given matching of subtasks to machines. The data relocation scheme is then extended to reduce intermachine data communication time in an HC environment with a given matching and scheduling of subtasks' computation where: 1) multiple subtasks' computations can be performed concurrently on different machines; 2) subtask computation stteps can be overlapped with other subtasks' communication steps for intermachine data transfers; and 3) machines in the HC suite are interconnected by a shared-bus type of network.

Quisiéramos mostrar nuestro más sincero agradecimiento a nuestros tutores de proyecto Guillermo Botella Juan y Carlos García Sánchez, ya que sin ellos hubiera sido imposible la realización del mismo. Nos han apoyado y motivado desde el primer momento consiguiendo que nos interesáramos cada día más por la temática del proyecto, al igual que han estado siempre a nuestra disposición para todo aquello que hemos necesitado. Por supuesto agradecer a nuestras familias, parejas y amigos el apoyo diario que nos han dado durante todos estos años hasta llegar aquí.

The Management System for Heterogeneous Networks (MSHN) is a large, distributed research software system project that began over 18 months ago. The primary goal of MSHN is to develop a framework within which nextgeneration resource management system (RMS) issues can be investigated. The initial MSHN design was developed using basic object-oriented design principles and the initial prototype was built with object-oriented technology (IDL, C++, Java, and CORBA). After building an initial proof-of-concept prototype, we looked to the standardized terms, symbols and diagrams of the Unified Modeling Language (UML) to explain the functionality of MSHN to new students and staff members of the development group, as well as interested colleagues outside of the group. As we learned more about the UML and applied it to MSHN in further detail, we found that this semi-formal method not only (i) helped us to communicate MSHN’s functionality to others, but also (ii) improved our design, helping us ...

Actors are computational entities that communicate with other actors and the environment by passing tokens via theirs input and output ports. Actors have state and parameters and when an actor is fired, input tokens are consumed tokens, output tokens are produced, and the internal state is updates. Actors are connected to form models or applications. It is highly desirable to be able to investigate properties, such as bounded memory and absence of deadlock, for such applications. One key for doing this is to have full knowledge of the production and consumption patterns of the actors. Another issue for actor applications is to generate code for embedded systems. To do so is necessary to have a good understanding of control flow and memory management. Caltrop is novel actor language designed to facilitate the extraction of needed properties in order handle those issues. It is a small domain specific language for implementation of actors. It is designed to be embedded in richer environment, providing data type, operations and function libraries.

This paper considers the contributions of one outstanding as well as controversial researcher, teacher and inventor from Spain during the Francos´s dictatorship. The paper presents an overview of the most important contributions made by Prof. Angel Gonzalez del Valle related with the electric networks/circuits theory and its application to analog computing/simulation, which was achieved from forties to sixties in the last century in Spain. A list of publications about those topics is presented.

With energy efficiency becoming a major concern in the HPC community, low-power alternatives such as heterogeneous Multi-Processor System-on-chip (MPSoC) are gaining popularity. These devices house several different kinds of accelerators on-chip including Digital Signal Processors (DSP) alongside CPU and GPU cores. Maximum performance from MPSoC devices can be achieved through the simultaneous use of all the processing elements. In addition to partitioning work across different processing elements being notoriously difficult on MPSoC platforms, achieving an energy-optimal partition is known to be non-trivial. The Keystone II Hawking (K2H) Platform is one such MPSoC from Texas Instruments that consists of 4 cache coherent ARM cores and 8 TI DSP cores that are not cache coherent. This MPSoC provides high floating-point performance with low power consumption and is also being used in the nCore BrownDwarf supercomputer. Recent work on the K2H MPSoC describes a novel hybrid work-stealing...

The augmented assembly language @Asm is proposed in order to transcend the fragmentation of architecture-specific dialects, to provide a unified framework for diverse processing paradigms as a universal assembly language and to function as a self-compiling bootstrap instrument adaptable to any processor system. The language augmentations include: flexible machine-language descriptions, general memory and data management directives, custom lexical identification through regular expressions, parsing facilities, generalized macroprocessing, flexible assembly control instructions, customizable encoding and code generation features, compiler-oriented abstraction mechanisms at the language level. The native abstraction augmentations enable expressive and concise high-level descriptions within assembly language for any present, emerging, or future systems.

We present a local search strategy to improve the mapping of a parallel job’s tasks to the MPI ranks of its parallel allocation in order to reduce network congestion and the job’s communication time. The goal is to reduce the number of network hops between communicating pairs of ranks. Our target is applications with a nearest-neighbor stencil communication pattern running on mesh systems with non-contiguous processor allocation, such as Cray XE and XK Systems. Using the miniGhost mini-app, which models the shock physics application CTH, we demonstrate that our strategy reduces application running time while also reducing the runtime variability. Keywords-Task mapping, stencil communication pattern, non-contiguous allocation, local search.

The widening disparity between processor execution speed and main memory latency continues to pose a fundamental challenge in computer architecture design. Memory hierarchies, particularly cache-based systems, have emerged as the principal mechanism for alleviating this bottleneck by exploiting program locality and optimizing data access patterns. Nevertheless, cache design is inherently constrained by competing performance objectives involving latency, miss rate, power consumption, silicon area, and implementation complexity. This paper presents a quantitative evaluation of memory hierarchy design with a focus on performance trade-offs in contemporary cache architectures. Using analytical performance models and architectural case studies, the study examines how cache parameters such as size, associativity, block organization, and multi-level hierarchy structure influence system performance. The findings underscore that optimal cache design requires a balanced consideration of multiple metrics rather than isolated optimization, particularly in modern processors operating under diverse workload and power constraints.

A frequently used method of clustering is a technique called kmeans clustering. The k-means algorithm consists of two steps: A map step, which is simple to execute on a GPU, and a reduce step, which is more problematic. Previous researchers have used a hybrid approach in which the map step is computed on the GPU and the reduce step is performed on the CPU. In this work, we present a new algorithm for irregular reductions and apply it to k-means such that the GPU executes both the map and reduce steps. We provide experimental comparisons using OpenCL. Our results show that our scheme is 3.2 times faster than the hybrid scheme for k = 10, an average 1.5 times faster when the number of clusters, k = 100 and on average equal for k = 400, on an ATI Radeon® HD 5870 (best speedup was 3.5 times) compared to the hybrid approach. In addition, we compare the GPU code with the standard OpenMP benchmark, MineBench. In that implementation, both the map and reduce steps are computed on the CPU. For large data sizes, the new GPU scheme shows great promise, with performance up to 35 times faster than MineBench on a four core Intel i7 CPU.

The sizes of GPU applications are rapidly growing. They are exhausting the compute and memory resources of a single GPU, and are demanding the move to multiple GPUs. However, the performance of these applications scales sub-linearly with GPU count because of the overhead of data movement across multiple GPUs. Moreover, a lack of hardware support for coherency exacerbates the problem because a programmer must either replicate the data across GPUs or fetch the remote data using high-overhead off-chip links. To address these problems, we propose a multi-GPU system with truly shared memory (MGPU-TSM), where the main memory is physically shared across all the GPUs. We eliminate remote accesses and avoid data replication using an MGPU-TSM system, which simplifies the memory hierarchy. Our preliminary analysis shows that MGPU-TSM with 4 GPUs performs, on average, 3.9x? better than the current best performing multi-GPU configuration for standard application benchmarks.

AbstractÐEfficient application scheduling is critical for achieving high performance in heterogeneous computing environments. The application scheduling problem has been shown to be NP-complete in general cases as well as in several restricted cases. Because of its key importance, this problem has been extensively studied and various algorithms have been proposed in the literature which are mainly for systems with homogeneous processors. Although there are a few algorithms in the literature for heterogeneous processors, they usually require significantly high scheduling costs and they may not deliver good quality schedules with lower costs. In this paper, we present two novel scheduling algorithms for a bounded number of heterogeneous processors with an objective to simultaneously meet high performance and fast scheduling time, which are called the Heterogeneous Earliest-Finish-Time (HEFT) algorithm and the Critical-Path-on-a-Processor (CPOP) algorithm. The HEFT algorithm selects the task with the highest upward rank value at each step and assigns the selected task to the processor, which minimizes its earliest finish time with an insertion-based approach. On the other hand, the CPOP algorithm uses the summation of upward and downward rank values for prioritizing tasks. Another difference is in the processor selection phase, which schedules the critical tasks onto the processor that minimizes the total execution time of the critical tasks. In order to provide a robust and unbiased comparison with the related work, a parametric graph generator was designed to generate weighted directed acyclic graphs with various characteristics. The comparison study, based on both randomly generated graphs and the graphs of some real applications, shows that our scheduling algorithms significantly surpass previous approaches in terms of both quality and cost of schedules, which are mainly presented with schedule length ratio, speedup, frequency of best results, and average scheduling time metrics.